室内定位中K-means聚类算法奇异值的优化处理

针对室内定位聚类算法中的奇异值出现较多的场景,按照以往聚类算法大多将其删除或替代为聚类平均值,这往往使得奇异值附近的定位误差陡增。研究采集阶段接入点(acess point,AP)端加入嵌入式滤波处理单元,采用格拉布斯(Grubbs)准则处理采集的信号以减少检测奇异值;然后在定位运算中改进了K-means聚类算法。首先根据模型函数鉴别运算中产生的奇异值,将奇异值线性化处理后由支持向量机(sport vector machine,SVM)对于奇异点进行分类;再将其进行K-means聚类划分。在不剔除奇异值的情况下,使得定位区域中的参考点合理利用,从而提高了整体累计误差的置信水平。研究中将剔除奇异值的K-means聚类算法作为比较对象,实验中采用美国Signal Hound公司的SA44B型频谱仪测量接收机组成传感器网络,可以使得K-means聚类算法的定位精度提高11.3%,证明在实际定位应用中是很有效的。

Optimal Processing of Singular Values of K-means Clustering Algorithm in Indoor Location

According to the indoor location clustering algorithm of the singular value are in accordance with the previous scene, most of the clustering algorithm to delete or replace the average value of the clustering, which often make the singular value on the positioning error near the acquisition access point in the study (Acess Point AP) end embedded filter processing unit, by Grab J (Grubbs) signal acquisition and processing of the standards in order to reduce the detection of singular value, and then in the positioning operation improved K-means clustering algorithm, based on model identification function in the operation of singular value, singular value after linearization by support vector machine (Sport Vector Machine, SVM) for the classification of singular points, and then its K-means clustering allows the location area of the reference point in eliminating singular values under the condition of reasonable utilization, thereby improve the overall cumulative error Confidence level. K-means clustering algorithm in eliminating singular values as the comparison object, type SA44B measuring receiver Signal Hound company using wireless sensor network, the method can make the positioning accuracy of the K-means clustering algorithm is increased by 11.3%, proved to be effective in the practical application of positioning.